2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro, <bir7@leland.Stanford.Edu>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
60 #include <linux/config.h>
62 #include <linux/smp_lock.h>
63 #include <linux/socket.h>
64 #include <linux/file.h>
65 #include <linux/net.h>
66 #include <linux/interrupt.h>
67 #include <linux/netdevice.h>
68 #include <linux/proc_fs.h>
69 #include <linux/wanrouter.h>
70 #include <linux/init.h>
71 #include <linux/poll.h>
72 #include <linux/cache.h>
73 #include <linux/module.h>
75 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
76 #include <linux/kmod.h>
79 #include <asm/uaccess.h>
81 #include <linux/inet.h>
87 #include <linux/netfilter.h>
89 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
90 static loff_t
sock_lseek(struct file
*file
, loff_t offset
, int whence
);
91 static ssize_t
sock_read(struct file
*file
, char *buf
,
92 size_t size
, loff_t
*ppos
);
93 static ssize_t
sock_write(struct file
*file
, const char *buf
,
94 size_t size
, loff_t
*ppos
);
95 static int sock_mmap(struct file
*file
, struct vm_area_struct
* vma
);
97 static int sock_close(struct inode
*inode
, struct file
*file
);
98 static unsigned int sock_poll(struct file
*file
,
99 struct poll_table_struct
*wait
);
100 static int sock_ioctl(struct inode
*inode
, struct file
*file
,
101 unsigned int cmd
, unsigned long arg
);
102 static int sock_fasync(int fd
, struct file
*filp
, int on
);
103 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
104 unsigned long count
, loff_t
*ppos
);
105 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
106 unsigned long count
, loff_t
*ppos
);
110 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
111 * in the operation structures but are done directly via the socketcall() multiplexor.
114 static struct file_operations socket_file_ops
= {
121 open
: sock_no_open
, /* special open code to disallow open via /proc */
129 * The protocol list. Each protocol is registered in here.
132 static struct net_proto_family
*net_families
[NPROTO
];
135 static atomic_t net_family_lockct
= ATOMIC_INIT(0);
136 static spinlock_t net_family_lock
= SPIN_LOCK_UNLOCKED
;
138 /* The strategy is: modifications net_family vector are short, do not
139 sleep and veeery rare, but read access should be free of any exclusive
143 static void net_family_write_lock(void)
145 spin_lock(&net_family_lock
);
146 while (atomic_read(&net_family_lockct
) != 0) {
147 spin_unlock(&net_family_lock
);
149 current
->policy
|= SCHED_YIELD
;
152 spin_lock(&net_family_lock
);
156 static __inline__
void net_family_write_unlock(void)
158 spin_unlock(&net_family_lock
);
161 static __inline__
void net_family_read_lock(void)
163 atomic_inc(&net_family_lockct
);
164 spin_unlock_wait(&net_family_lock
);
167 static __inline__
void net_family_read_unlock(void)
169 atomic_dec(&net_family_lockct
);
173 #define net_family_write_lock() do { } while(0)
174 #define net_family_write_unlock() do { } while(0)
175 #define net_family_read_lock() do { } while(0)
176 #define net_family_read_unlock() do { } while(0)
181 * Statistics counters of the socket lists
186 char __pad
[SMP_CACHE_BYTES
];
187 } sockets_in_use
[NR_CPUS
] __cacheline_aligned
= {{0}};
190 * Support routines. Move socket addresses back and forth across the kernel/user
191 * divide and look after the messy bits.
194 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
195 16 for IP, 16 for IPX,
198 must be at least one bigger than
199 the AF_UNIX size (see net/unix/af_unix.c
204 * move_addr_to_kernel - copy a socket address into kernel space
205 * @uaddr: Address in user space
206 * @kaddr: Address in kernel space
207 * @ulen: Length in user space
209 * The address is copied into kernel space. If the provided address is
210 * too long an error code of -EINVAL is returned. If the copy gives
211 * invalid addresses -EFAULT is returned. On a success 0 is returned.
214 int move_addr_to_kernel(void *uaddr
, int ulen
, void *kaddr
)
216 if(ulen
<0||ulen
>MAX_SOCK_ADDR
)
220 if(copy_from_user(kaddr
,uaddr
,ulen
))
226 * move_addr_to_user - copy an address to user space
227 * @kaddr: kernel space address
228 * @klen: length of address in kernel
229 * @uaddr: user space address
230 * @ulen: pointer to user length field
232 * The value pointed to by ulen on entry is the buffer length available.
233 * This is overwritten with the buffer space used. -EINVAL is returned
234 * if an overlong buffer is specified or a negative buffer size. -EFAULT
235 * is returned if either the buffer or the length field are not
237 * After copying the data up to the limit the user specifies, the true
238 * length of the data is written over the length limit the user
239 * specified. Zero is returned for a success.
242 int move_addr_to_user(void *kaddr
, int klen
, void *uaddr
, int *ulen
)
247 if((err
=get_user(len
, ulen
)))
251 if(len
<0 || len
> MAX_SOCK_ADDR
)
255 if(copy_to_user(uaddr
,kaddr
,len
))
259 * "fromlen shall refer to the value before truncation.."
262 return __put_user(klen
, ulen
);
265 #define SOCKFS_MAGIC 0x534F434B
266 static int sockfs_statfs(struct super_block
*sb
, struct statfs
*buf
)
268 buf
->f_type
= SOCKFS_MAGIC
;
270 buf
->f_namelen
= 255;
274 static struct super_operations sockfs_ops
= {
275 statfs
: sockfs_statfs
,
278 static struct super_block
* sockfs_read_super(struct super_block
*sb
, void *data
, int silent
)
280 struct inode
*root
= get_empty_inode();
283 root
->i_mode
= S_IFDIR
| S_IRUSR
| S_IWUSR
;
284 root
->i_uid
= root
->i_gid
= 0;
285 root
->i_atime
= root
->i_mtime
= root
->i_ctime
= CURRENT_TIME
;
287 root
->i_dev
= sb
->s_dev
;
288 sb
->s_blocksize
= 1024;
289 sb
->s_blocksize_bits
= 10;
290 sb
->s_magic
= SOCKFS_MAGIC
;
291 sb
->s_op
= &sockfs_ops
;
292 sb
->s_root
= d_alloc(NULL
, &(const struct qstr
) { "socket:", 7, 0 });
297 sb
->s_root
->d_sb
= sb
;
298 sb
->s_root
->d_parent
= sb
->s_root
;
299 d_instantiate(sb
->s_root
, root
);
303 static struct vfsmount
*sock_mnt
;
304 static DECLARE_FSTYPE(sock_fs_type
, "sockfs", sockfs_read_super
,
305 FS_NOMOUNT
|FS_SINGLE
);
306 static int sockfs_delete_dentry(struct dentry
*dentry
)
310 static struct dentry_operations sockfs_dentry_operations
= {
311 d_delete
: sockfs_delete_dentry
,
315 * Obtains the first available file descriptor and sets it up for use.
317 * This functions creates file structure and maps it to fd space
318 * of current process. On success it returns file descriptor
319 * and file struct implicitly stored in sock->file.
320 * Note that another thread may close file descriptor before we return
321 * from this function. We use the fact that now we do not refer
322 * to socket after mapping. If one day we will need it, this
323 * function will inincrement ref. count on file by 1.
325 * In any case returned fd MAY BE not valid!
326 * This race condition is inavoidable
327 * with shared fd spaces, we cannot solve is inside kernel,
328 * but we take care of internal coherence yet.
331 static int sock_map_fd(struct socket
*sock
)
338 * Find a file descriptor suitable for return to the user.
341 fd
= get_unused_fd();
343 struct file
*file
= get_empty_filp();
351 sprintf(name
, "[%lu]", sock
->inode
->i_ino
);
353 this.len
= strlen(name
);
354 this.hash
= sock
->inode
->i_ino
;
356 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
357 if (!file
->f_dentry
) {
363 file
->f_dentry
->d_op
= &sockfs_dentry_operations
;
364 d_add(file
->f_dentry
, sock
->inode
);
365 file
->f_vfsmnt
= mntget(sock_mnt
);
368 file
->f_op
= sock
->inode
->i_fop
= &socket_file_ops
;
370 file
->f_flags
= O_RDWR
;
372 fd_install(fd
, file
);
379 extern __inline__
struct socket
*socki_lookup(struct inode
*inode
)
381 return &inode
->u
.socket_i
;
385 * sockfd_lookup - Go from a file number to its socket slot
387 * @err: pointer to an error code return
389 * The file handle passed in is locked and the socket it is bound
390 * too is returned. If an error occurs the err pointer is overwritten
391 * with a negative errno code and NULL is returned. The function checks
392 * for both invalid handles and passing a handle which is not a socket.
394 * On a success the socket object pointer is returned.
397 struct socket
*sockfd_lookup(int fd
, int *err
)
403 if (!(file
= fget(fd
)))
409 inode
= file
->f_dentry
->d_inode
;
410 if (!inode
|| !inode
->i_sock
|| !(sock
= socki_lookup(inode
)))
417 if (sock
->file
!= file
) {
418 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
424 extern __inline__
void sockfd_put(struct socket
*sock
)
430 * sock_alloc - allocate a socket
432 * Allocate a new inode and socket object. The two are bound together
433 * and initialised. The socket is then returned. If we are out of inodes
437 struct socket
*sock_alloc(void)
439 struct inode
* inode
;
440 struct socket
* sock
;
442 inode
= get_empty_inode();
446 sock
= socki_lookup(inode
);
448 inode
->i_mode
= S_IFSOCK
|S_IRWXUGO
;
450 inode
->i_uid
= current
->fsuid
;
451 inode
->i_gid
= current
->fsgid
;
454 init_waitqueue_head(&sock
->wait
);
455 sock
->fasync_list
= NULL
;
456 sock
->state
= SS_UNCONNECTED
;
462 sockets_in_use
[smp_processor_id()].counter
++;
467 * In theory you can't get an open on this inode, but /proc provides
468 * a back door. Remember to keep it shut otherwise you'll let the
469 * creepy crawlies in.
472 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
478 * sock_release - close a socket
479 * @sock: socket to close
481 * The socket is released from the protocol stack if it has a release
482 * callback, and the inode is then released if the socket is bound to
483 * an inode not a file.
486 void sock_release(struct socket
*sock
)
489 sock
->ops
->release(sock
);
491 if (sock
->fasync_list
)
492 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
494 sockets_in_use
[smp_processor_id()].counter
--;
502 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, int size
)
505 struct scm_cookie scm
;
507 err
= scm_send(sock
, msg
, &scm
);
509 err
= sock
->ops
->sendmsg(sock
, msg
, size
, &scm
);
515 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int size
, int flags
)
517 struct scm_cookie scm
;
519 memset(&scm
, 0, sizeof(scm
));
521 size
= sock
->ops
->recvmsg(sock
, msg
, size
, flags
, &scm
);
523 scm_recv(sock
, msg
, &scm
, flags
);
530 * Sockets are not seekable.
533 static loff_t
sock_lseek(struct file
*file
, loff_t offset
, int whence
)
539 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
540 * area ubuf...ubuf+size-1 is writable before asking the protocol.
543 static ssize_t
sock_read(struct file
*file
, char *ubuf
,
544 size_t size
, loff_t
*ppos
)
551 if (ppos
!= &file
->f_pos
)
553 if (size
==0) /* Match SYS5 behaviour */
556 sock
= socki_lookup(file
->f_dentry
->d_inode
);
562 msg
.msg_control
=NULL
;
563 msg
.msg_controllen
=0;
566 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
568 return sock_recvmsg(sock
, &msg
, size
, flags
);
573 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
574 * is readable by the user process.
577 static ssize_t
sock_write(struct file
*file
, const char *ubuf
,
578 size_t size
, loff_t
*ppos
)
584 if (ppos
!= &file
->f_pos
)
586 if(size
==0) /* Match SYS5 behaviour */
589 sock
= socki_lookup(file
->f_dentry
->d_inode
);
595 msg
.msg_control
=NULL
;
596 msg
.msg_controllen
=0;
597 msg
.msg_flags
=!(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
598 if (sock
->type
== SOCK_SEQPACKET
)
599 msg
.msg_flags
|= MSG_EOR
;
600 iov
.iov_base
=(void *)ubuf
;
603 return sock_sendmsg(sock
, &msg
, size
);
606 int sock_readv_writev(int type
, struct inode
* inode
, struct file
* file
,
607 const struct iovec
* iov
, long count
, long size
)
612 sock
= socki_lookup(inode
);
616 msg
.msg_control
= NULL
;
617 msg
.msg_controllen
= 0;
618 msg
.msg_iov
= (struct iovec
*) iov
;
619 msg
.msg_iovlen
= count
;
620 msg
.msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
622 /* read() does a VERIFY_WRITE */
623 if (type
== VERIFY_WRITE
)
624 return sock_recvmsg(sock
, &msg
, size
, msg
.msg_flags
);
626 if (sock
->type
== SOCK_SEQPACKET
)
627 msg
.msg_flags
|= MSG_EOR
;
629 return sock_sendmsg(sock
, &msg
, size
);
632 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
633 unsigned long count
, loff_t
*ppos
)
637 for (i
= 0 ; i
< count
; i
++)
638 tot_len
+= vector
[i
].iov_len
;
639 return sock_readv_writev(VERIFY_WRITE
, file
->f_dentry
->d_inode
,
640 file
, vector
, count
, tot_len
);
643 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
644 unsigned long count
, loff_t
*ppos
)
648 for (i
= 0 ; i
< count
; i
++)
649 tot_len
+= vector
[i
].iov_len
;
650 return sock_readv_writev(VERIFY_READ
, file
->f_dentry
->d_inode
,
651 file
, vector
, count
, tot_len
);
655 * With an ioctl arg may well be a user mode pointer, but we don't know what to do
656 * with it - that's up to the protocol still.
659 int sock_ioctl(struct inode
*inode
, struct file
*file
, unsigned int cmd
,
666 sock
= socki_lookup(inode
);
667 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
674 /* No kernel lock held - perfect */
675 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
680 * We can't return errors to poll, so it's either yes or no.
682 sock
= socki_lookup(file
->f_dentry
->d_inode
);
683 return sock
->ops
->poll(file
, sock
, wait
);
686 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
688 struct socket
*sock
= socki_lookup(file
->f_dentry
->d_inode
);
690 return sock
->ops
->mmap(file
, sock
, vma
);
693 int sock_close(struct inode
*inode
, struct file
*filp
)
696 * It was possible the inode is NULL we were
697 * closing an unfinished socket.
702 printk(KERN_DEBUG
"sock_close: NULL inode\n");
705 sock_fasync(-1, filp
, 0);
706 sock_release(socki_lookup(inode
));
711 * Update the socket async list
713 * Fasync_list locking strategy.
715 * 1. fasync_list is modified only under process context socket lock
716 * i.e. under semaphore.
717 * 2. fasync_list is used under read_lock(&sk->callback_lock)
718 * or under socket lock.
719 * 3. fasync_list can be used from softirq context, so that
720 * modification under socket lock have to be enhanced with
721 * write_lock_bh(&sk->callback_lock).
725 static int sock_fasync(int fd
, struct file
*filp
, int on
)
727 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
733 fna
=(struct fasync_struct
*)kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
739 sock
= socki_lookup(filp
->f_dentry
->d_inode
);
741 if ((sk
=sock
->sk
) == NULL
)
746 prev
=&(sock
->fasync_list
);
748 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
749 if (fa
->fa_file
==filp
)
756 write_lock_bh(&sk
->callback_lock
);
758 write_unlock_bh(&sk
->callback_lock
);
760 kfree_s(fna
,sizeof(struct fasync_struct
));
765 fna
->magic
=FASYNC_MAGIC
;
766 fna
->fa_next
=sock
->fasync_list
;
767 write_lock_bh(&sk
->callback_lock
);
768 sock
->fasync_list
=fna
;
769 write_unlock_bh(&sk
->callback_lock
);
775 write_lock_bh(&sk
->callback_lock
);
777 write_unlock_bh(&sk
->callback_lock
);
778 kfree_s(fa
,sizeof(struct fasync_struct
));
783 release_sock(sock
->sk
);
787 /* This function may be called only under socket lock or callback_lock */
789 int sock_wake_async(struct socket
*sock
, int how
, int band
)
791 if (!sock
|| !sock
->fasync_list
)
797 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
801 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
806 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
809 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
815 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
821 * Check protocol is in range
823 if(family
<0 || family
>=NPROTO
)
828 This uglymoron is moved from INET layer to here to avoid
829 deadlock in module load.
831 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
835 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n", current
->comm
);
840 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
841 /* Attempt to load a protocol module if the find failed.
843 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
844 * requested real, full-featured networking support upon configuration.
845 * Otherwise module support will break!
847 if (net_families
[family
]==NULL
)
849 char module_name
[30];
850 sprintf(module_name
,"net-pf-%d",family
);
851 request_module(module_name
);
855 net_family_read_lock();
856 if (net_families
[family
] == NULL
) {
862 * Allocate the socket and allow the family to set things up. if
863 * the protocol is 0, the family is instructed to select an appropriate
867 if (!(sock
= sock_alloc()))
869 printk(KERN_WARNING
"socket: no more sockets\n");
870 i
= -ENFILE
; /* Not exactly a match, but its the
871 closest posix thing */
877 if ((i
= net_families
[family
]->create(sock
, protocol
)) < 0)
886 net_family_read_unlock();
890 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
895 retval
= sock_create(family
, type
, protocol
, &sock
);
899 retval
= sock_map_fd(sock
);
904 /* It may be already another descriptor 8) Not kernel problem. */
913 * Create a pair of connected sockets.
916 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int usockvec
[2])
918 struct socket
*sock1
, *sock2
;
922 * Obtain the first socket and check if the underlying protocol
923 * supports the socketpair call.
926 err
= sock_create(family
, type
, protocol
, &sock1
);
930 err
= sock_create(family
, type
, protocol
, &sock2
);
934 err
= sock1
->ops
->socketpair(sock1
, sock2
);
936 goto out_release_both
;
940 err
= sock_map_fd(sock1
);
942 goto out_release_both
;
945 err
= sock_map_fd(sock2
);
950 /* fd1 and fd2 may be already another descriptors.
951 * Not kernel problem.
954 err
= put_user(fd1
, &usockvec
[0]);
956 err
= put_user(fd2
, &usockvec
[1]);
979 * Bind a name to a socket. Nothing much to do here since it's
980 * the protocol's responsibility to handle the local address.
982 * We move the socket address to kernel space before we call
983 * the protocol layer (having also checked the address is ok).
986 asmlinkage
long sys_bind(int fd
, struct sockaddr
*umyaddr
, int addrlen
)
989 char address
[MAX_SOCK_ADDR
];
992 if((sock
= sockfd_lookup(fd
,&err
))!=NULL
)
994 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0)
995 err
= sock
->ops
->bind(sock
, (struct sockaddr
*)address
, addrlen
);
1003 * Perform a listen. Basically, we allow the protocol to do anything
1004 * necessary for a listen, and if that works, we mark the socket as
1005 * ready for listening.
1008 asmlinkage
long sys_listen(int fd
, int backlog
)
1010 struct socket
*sock
;
1013 if ((sock
= sockfd_lookup(fd
, &err
)) != NULL
) {
1014 if ((unsigned) backlog
> SOMAXCONN
)
1015 backlog
= SOMAXCONN
;
1016 err
=sock
->ops
->listen(sock
, backlog
);
1024 * For accept, we attempt to create a new socket, set up the link
1025 * with the client, wake up the client, then return the new
1026 * connected fd. We collect the address of the connector in kernel
1027 * space and move it to user at the very end. This is unclean because
1028 * we open the socket then return an error.
1030 * 1003.1g adds the ability to recvmsg() to query connection pending
1031 * status to recvmsg. We need to add that support in a way thats
1032 * clean when we restucture accept also.
1035 asmlinkage
long sys_accept(int fd
, struct sockaddr
*upeer_sockaddr
, int *upeer_addrlen
)
1037 struct socket
*sock
, *newsock
;
1039 char address
[MAX_SOCK_ADDR
];
1041 sock
= sockfd_lookup(fd
, &err
);
1046 if (!(newsock
= sock_alloc()))
1049 newsock
->type
= sock
->type
;
1050 newsock
->ops
= sock
->ops
;
1052 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1056 if (upeer_sockaddr
) {
1057 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1058 err
= -ECONNABORTED
;
1061 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1066 /* File flags are not inherited via accept() unlike another OSes. */
1068 if ((err
= sock_map_fd(newsock
)) < 0)
1077 sock_release(newsock
);
1083 * Attempt to connect to a socket with the server address. The address
1084 * is in user space so we verify it is OK and move it to kernel space.
1086 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1089 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1090 * other SEQPACKET protocols that take time to connect() as it doesn't
1091 * include the -EINPROGRESS status for such sockets.
1094 asmlinkage
long sys_connect(int fd
, struct sockaddr
*uservaddr
, int addrlen
)
1096 struct socket
*sock
;
1097 char address
[MAX_SOCK_ADDR
];
1100 sock
= sockfd_lookup(fd
, &err
);
1103 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1106 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1107 sock
->file
->f_flags
);
1115 * Get the local address ('name') of a socket object. Move the obtained
1116 * name to user space.
1119 asmlinkage
long sys_getsockname(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
1121 struct socket
*sock
;
1122 char address
[MAX_SOCK_ADDR
];
1125 sock
= sockfd_lookup(fd
, &err
);
1128 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1131 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1140 * Get the remote address ('name') of a socket object. Move the obtained
1141 * name to user space.
1144 asmlinkage
long sys_getpeername(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
1146 struct socket
*sock
;
1147 char address
[MAX_SOCK_ADDR
];
1150 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1152 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1154 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1161 * Send a datagram to a given address. We move the address into kernel
1162 * space and check the user space data area is readable before invoking
1166 asmlinkage
long sys_sendto(int fd
, void * buff
, size_t len
, unsigned flags
,
1167 struct sockaddr
*addr
, int addr_len
)
1169 struct socket
*sock
;
1170 char address
[MAX_SOCK_ADDR
];
1175 sock
= sockfd_lookup(fd
, &err
);
1183 msg
.msg_control
=NULL
;
1184 msg
.msg_controllen
=0;
1185 msg
.msg_namelen
=addr_len
;
1188 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1191 msg
.msg_name
=address
;
1193 if (sock
->file
->f_flags
& O_NONBLOCK
)
1194 flags
|= MSG_DONTWAIT
;
1195 msg
.msg_flags
= flags
;
1196 err
= sock_sendmsg(sock
, &msg
, len
);
1205 * Send a datagram down a socket.
1208 asmlinkage
long sys_send(int fd
, void * buff
, size_t len
, unsigned flags
)
1210 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1214 * Receive a frame from the socket and optionally record the address of the
1215 * sender. We verify the buffers are writable and if needed move the
1216 * sender address from kernel to user space.
1219 asmlinkage
long sys_recvfrom(int fd
, void * ubuf
, size_t size
, unsigned flags
,
1220 struct sockaddr
*addr
, int *addr_len
)
1222 struct socket
*sock
;
1225 char address
[MAX_SOCK_ADDR
];
1228 sock
= sockfd_lookup(fd
, &err
);
1232 msg
.msg_control
=NULL
;
1233 msg
.msg_controllen
=0;
1238 msg
.msg_name
=address
;
1239 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1240 if (sock
->file
->f_flags
& O_NONBLOCK
)
1241 flags
|= MSG_DONTWAIT
;
1242 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1244 if(err
>= 0 && addr
!= NULL
&& msg
.msg_namelen
)
1246 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1256 * Receive a datagram from a socket.
1259 asmlinkage
long sys_recv(int fd
, void * ubuf
, size_t size
, unsigned flags
)
1261 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1265 * Set a socket option. Because we don't know the option lengths we have
1266 * to pass the user mode parameter for the protocols to sort out.
1269 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char *optval
, int optlen
)
1272 struct socket
*sock
;
1274 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1276 if (level
== SOL_SOCKET
)
1277 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1279 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1286 * Get a socket option. Because we don't know the option lengths we have
1287 * to pass a user mode parameter for the protocols to sort out.
1290 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char *optval
, int *optlen
)
1293 struct socket
*sock
;
1295 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1297 if (level
== SOL_SOCKET
)
1298 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1300 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1308 * Shutdown a socket.
1311 asmlinkage
long sys_shutdown(int fd
, int how
)
1314 struct socket
*sock
;
1316 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1318 err
=sock
->ops
->shutdown(sock
, how
);
1325 * BSD sendmsg interface
1328 asmlinkage
long sys_sendmsg(int fd
, struct msghdr
*msg
, unsigned flags
)
1330 struct socket
*sock
;
1331 char address
[MAX_SOCK_ADDR
];
1332 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1333 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]; /* 20 is size of ipv6_pktinfo */
1334 unsigned char *ctl_buf
= ctl
;
1335 struct msghdr msg_sys
;
1336 int err
, ctl_len
, iov_size
, total_len
;
1339 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1342 sock
= sockfd_lookup(fd
, &err
);
1346 /* do not move before msg_sys is valid */
1348 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1351 /* Check whether to allocate the iovec area*/
1353 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1354 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1355 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1360 /* This will also move the address data into kernel space */
1361 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1368 if (msg_sys
.msg_controllen
> INT_MAX
)
1370 ctl_len
= msg_sys
.msg_controllen
;
1373 if (ctl_len
> sizeof(ctl
))
1376 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1377 if (ctl_buf
== NULL
)
1381 if (copy_from_user(ctl_buf
, msg_sys
.msg_control
, ctl_len
))
1383 msg_sys
.msg_control
= ctl_buf
;
1385 msg_sys
.msg_flags
= flags
;
1387 if (sock
->file
->f_flags
& O_NONBLOCK
)
1388 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1389 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1393 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1395 if (iov
!= iovstack
)
1396 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1404 * BSD recvmsg interface
1407 asmlinkage
long sys_recvmsg(int fd
, struct msghdr
*msg
, unsigned int flags
)
1409 struct socket
*sock
;
1410 struct iovec iovstack
[UIO_FASTIOV
];
1411 struct iovec
*iov
=iovstack
;
1412 struct msghdr msg_sys
;
1413 unsigned long cmsg_ptr
;
1414 int err
, iov_size
, total_len
, len
;
1416 /* kernel mode address */
1417 char addr
[MAX_SOCK_ADDR
];
1419 /* user mode address pointers */
1420 struct sockaddr
*uaddr
;
1424 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1427 sock
= sockfd_lookup(fd
, &err
);
1432 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1435 /* Check whether to allocate the iovec area*/
1437 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1438 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1439 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1445 * Save the user-mode address (verify_iovec will change the
1446 * kernel msghdr to use the kernel address space)
1449 uaddr
= msg_sys
.msg_name
;
1450 uaddr_len
= &msg
->msg_namelen
;
1451 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1456 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1457 msg_sys
.msg_flags
= 0;
1459 if (sock
->file
->f_flags
& O_NONBLOCK
)
1460 flags
|= MSG_DONTWAIT
;
1461 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1466 if (uaddr
!= NULL
&& msg_sys
.msg_namelen
) {
1467 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1471 err
= __put_user(msg_sys
.msg_flags
, &msg
->msg_flags
);
1474 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1475 &msg
->msg_controllen
);
1481 if (iov
!= iovstack
)
1482 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1491 * Perform a file control on a socket file descriptor.
1493 * Doesn't aquire a fd lock, because no network fcntl
1494 * function sleeps currently.
1497 int sock_fcntl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
1499 struct socket
*sock
;
1501 sock
= socki_lookup (filp
->f_dentry
->d_inode
);
1502 if (sock
&& sock
->ops
)
1503 return sock_no_fcntl(sock
, cmd
, arg
);
1507 /* Argument list sizes for sys_socketcall */
1508 #define AL(x) ((x) * sizeof(unsigned long))
1509 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1510 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1511 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1515 * System call vectors.
1517 * Argument checking cleaned up. Saved 20% in size.
1518 * This function doesn't need to set the kernel lock because
1519 * it is set by the callees.
1522 asmlinkage
long sys_socketcall(int call
, unsigned long *args
)
1525 unsigned long a0
,a1
;
1528 if(call
<1||call
>SYS_RECVMSG
)
1531 /* copy_from_user should be SMP safe. */
1532 if (copy_from_user(a
, args
, nargs
[call
]))
1541 err
= sys_socket(a0
,a1
,a
[2]);
1544 err
= sys_bind(a0
,(struct sockaddr
*)a1
, a
[2]);
1547 err
= sys_connect(a0
, (struct sockaddr
*)a1
, a
[2]);
1550 err
= sys_listen(a0
,a1
);
1553 err
= sys_accept(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1555 case SYS_GETSOCKNAME
:
1556 err
= sys_getsockname(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1558 case SYS_GETPEERNAME
:
1559 err
= sys_getpeername(a0
, (struct sockaddr
*)a1
, (int *)a
[2]);
1561 case SYS_SOCKETPAIR
:
1562 err
= sys_socketpair(a0
,a1
, a
[2], (int *)a
[3]);
1565 err
= sys_send(a0
, (void *)a1
, a
[2], a
[3]);
1568 err
= sys_sendto(a0
,(void *)a1
, a
[2], a
[3],
1569 (struct sockaddr
*)a
[4], a
[5]);
1572 err
= sys_recv(a0
, (void *)a1
, a
[2], a
[3]);
1575 err
= sys_recvfrom(a0
, (void *)a1
, a
[2], a
[3],
1576 (struct sockaddr
*)a
[4], (int *)a
[5]);
1579 err
= sys_shutdown(a0
,a1
);
1581 case SYS_SETSOCKOPT
:
1582 err
= sys_setsockopt(a0
, a1
, a
[2], (char *)a
[3], a
[4]);
1584 case SYS_GETSOCKOPT
:
1585 err
= sys_getsockopt(a0
, a1
, a
[2], (char *)a
[3], (int *)a
[4]);
1588 err
= sys_sendmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1591 err
= sys_recvmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1601 * This function is called by a protocol handler that wants to
1602 * advertise its address family, and have it linked into the
1606 int sock_register(struct net_proto_family
*ops
)
1610 if (ops
->family
>= NPROTO
) {
1611 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
1614 net_family_write_lock();
1616 if (net_families
[ops
->family
] == NULL
) {
1617 net_families
[ops
->family
]=ops
;
1620 net_family_write_unlock();
1625 * This function is called by a protocol handler that wants to
1626 * remove its address family, and have it unlinked from the
1630 int sock_unregister(int family
)
1632 if (family
< 0 || family
>= NPROTO
)
1635 net_family_write_lock();
1636 net_families
[family
]=NULL
;
1637 net_family_write_unlock();
1641 void __init
proto_init(void)
1643 extern struct net_proto protocols
[]; /* Network protocols */
1644 struct net_proto
*pro
;
1646 /* Kick all configured protocols. */
1648 while (pro
->name
!= NULL
)
1650 (*pro
->init_func
)(pro
);
1653 /* We're all done... */
1656 extern void sk_init(void);
1658 #ifdef CONFIG_BRIDGE
1659 extern int br_init(void);
1662 #ifdef CONFIG_WAN_ROUTER
1663 extern void wanrouter_init(void);
1666 void __init
sock_init(void)
1670 printk(KERN_INFO
"Linux NET4.0 for Linux 2.3\n");
1671 printk(KERN_INFO
"Based upon Swansea University Computer Society NET3.039\n");
1674 * Initialize all address (protocol) families.
1677 for (i
= 0; i
< NPROTO
; i
++)
1678 net_families
[i
] = NULL
;
1681 * Initialize sock SLAB cache.
1688 * Initialize skbuff SLAB cache
1694 * Ethernet bridge layer.
1697 #ifdef CONFIG_BRIDGE
1705 #ifdef CONFIG_WAN_ROUTER
1710 * Initialize the protocols module.
1716 * The netlink device handler may be needed early.
1719 #ifdef CONFIG_RTNETLINK
1722 #ifdef CONFIG_NETLINK_DEV
1725 #ifdef CONFIG_NETFILTER
1728 register_filesystem(&sock_fs_type
);
1729 sock_mnt
= kern_mount(&sock_fs_type
);
1732 int socket_get_info(char *buffer
, char **start
, off_t offset
, int length
)
1737 for (cpu
=0; cpu
<smp_num_cpus
; cpu
++)
1738 counter
+= sockets_in_use
[cpu_logical_map(cpu
)].counter
;
1740 /* It can be negative, by the way. 8) */
1744 len
= sprintf(buffer
, "sockets: used %d\n", counter
);
1750 *start
= buffer
+ offset
;